Rapid Detection Of Germination Rate Of Multi-category Corn Seeds Based On Near Infrared Spectroscop

Choosing suitable arable land to plant high vigor corn seeds can significantly improve corn emergence rate and increase corn yield.Using traditional methods to identify corn seed varieties and detect seed vigor has shortcomings such as long test period,cumbersome operation process,high professional requirements for test personnel,and personal subjectivity in the test process.Therefore,this study used near-infrared spectroscopy to achieve rapid identification of corn varieties and rapid detection of seed vigor.The following results were achieved:（1）A Support Vector Machine（SVM）classification model for corn seeds based on Bayesian optimization was established.Taking 7 types of corn seeds as the research objects,the near-infrared spectral data of the seeds were collected and processed by Standard Normal Variate transformation（SNV）,and then Principal Component Analysis（PCA）was used to reduce the dimension of the spectral data.Randomly divide the sample data into training set and test set according to the ratio of 6:1,and use Grid Search（GS）,Genetic Algorithm（GA）and Bayesian Optimization Algorithm（BOA）respectively.The three methods optimize the two important parameters of the SVM model,the penalty factor C and the radial basis kernel function parameterγ,and establish the SVM classification model.Comparing the SVM models based on the three optimization algorithms,it is found that the performance of the SVM classification model optimized by BOA is significantly improved compared with the other two optimization algorithms,and the recognition accuracy and f₁ value on the test set can reach100%.It shows that the parameters of the SVM model optimized by BOA are the global optimal parameters,and the parameters obtained by the other two optimization algorithms make the model fall into the local optimum,resulting in poor model performance.The BOA-based SVM classification method is an effective method for identification of corn varieties.（2）The prediction model of corn seed germination rate was established.The corn seeds were artificially aged in different periods,and 8 gradients of artificial aging samples were obtained,including 0d,1d,2d,3d,4d,5d,6d,and 7d.Spectral data of aged samples were collected and germination tests were performed.Monte Carlo Cross Validation（MCCV）was used to remove abnormal samples,and the Successive Projections Algorithm（SPA）and Principal Component Analysis（PCA）were used for dimensionality reduction,and the effects of different wavelength points and the number of principal components on the performance of the model were discussed.Use the random partition method to divide and model the data multiple times to compare the stability and generalization performance of the model.Three regression models,Ridge Regression（RR）,Support Vector Regression（SVR）and Bayesian Ridge Regression（BRR）,were used to predict the germination rate of corn seeds.The results show that the GS-based SVR model performs poorly.The coefficient of determination R² of the BRR model and the RR model on the prediction set can both reach more than 0.93,and the mean absolute error MAE and the root mean square error RMSE are both less than 0.1.The determination coefficient R² of the BRR model on the prediction set can reach 0.9509,and its performance is better than that of RR.The experimental results show that the performance of BRR model is better than that of SVR and RR,and it is an effective method for rapid detection of corn seed germination rate.In this study,the effects of GS,GA and BOA on the performance of the SVM classification model were investigated,and a corn variety identification model was established.Through the germination test of corn seeds,three models of RR,SVR and BRR were established,and the model performance of the three models on the prediction set was compared,and the prediction model of the germination rate of corn seeds was established.It provides a method and theoretical basis for the rapid classification of corn varieties and the detection of seed vigor,and also provides a reference for the classification,origin identification and vigor detection of other agricultural products based on near-infrared spectroscopy.